Re: [PATCH -mm] swsusp: userland interface (rev 2)

[Kay Sievers <kay.sievers@vrfy.org>]

On Wed, Jan 25, 2006 at 09:09:27PM -0500, Jim Crilly wrote:
> On 01/24/06 11:44:37PM +0100, Pavel Machek wrote:
> > On Út 24-01-06 17:38:34, Dave Jones wrote:
> > >  > We'll of course try to get the interface right at the first
> > >  > try. OTOH... if wrong interface is in kernel for a month, I do not
> > >  > think it is reasonable to keep supporting that wrong interface for a
> > >  > year before it can be removed. One month of warning should be fair in
> > >  > such case...
> > > 
> > > Users want to be able to boot between different kernels.
> > > Tying functionality to specific versions of userspace completely
> > > screws them over.
> > 
> > Well, by the time we have any _users_ interface should be
> > stable. Actually I believe interface will be stable from day 0, but...
> >
> I'm sure gregkh thought the same thing with about sysfs and udev and we've
> seen how well that's worked out...

Well, that was just an unfortunate "bug".

Declaring interfaces "stable" makes as much sense as all the other
tries to define crazy enterprise "standards" nobody follows in real
world.

In a developing environment, interfaces _become_ stable and don't get
_declared_ by anybody as such. We are not talking about syscall
interfaces or things that are simple enough to be kept stable, if you
cross a certain level of complexity, you just can't apply these rules
anymore.

Interfaces mature over the time they get used. Only the _use_ of it
collects the needed information to form the model behind it. They get
improved up to the point that changing the interface causes more
pain than it's worth this change. Then an interface has _become_ "stable"
cause it makes sense at that point.

"by the time we have any _users_ interface should be stable", that's
such a nonsense. If you don't have any user, you don't know if this
interface works at all and only if it gets used you get the needed
feedback to improve it.

Kay
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/